Routine Blood-Based Parameters Associated with Invasive Cervical Cancer Versus High-Grade Cervical Intraepithelial Neoplasia: Development of a Retrospective Diagnostic Prediction Model.

Background/Objectives: Distinguishing invasive cervical cancer from high-grade cervical intraepithelial neoplasia (CIN) before histopathology remains difficult. We evaluated routine blood-based parameters for this distinction and developed and internally validated a parsimonious prediction model, testing its incremental value over age. Methods: In this retrospective diagnostic accuracy study with two-gate (case-control) sampling, we analyzed 137 unique patients with histopathologically confirmed cervical cancer and 238 with high-grade CIN after duplicate removal. A four-variable model (age, CEA, neutrophil-to-lymphocyte ratio (NLR), and hemoglobin) was developed and internally validated by bootstrap optimism correction, calibration, and decision-curve analysis, with age-matched and subgroup sensitivity analyses. Results: The strongest single discriminators were age (AUC 0.826), C-reactive protein (0.805), and albumin (0.772). The continuous model showed an apparent AUC of 0.859 (optimism-corrected 0.853; calibration slope 0.962; Brier 0.136). The gain over age alone was statistically significant but modest (ΔAUC +0.027; DeLong p = 0.015). After 1:1 age matching, age no longer discriminated (AUC 0.49), whereas the model retained moderate discrimination (AUC 0.686), with NLR independent. Performance was lower for CIN 3 versus early-stage cancer (AUC 0.759). Conclusions: The model may provide adjunctive risk-stratification information, but clinical implementation requires external validation in prospectively assembled cohorts; the findings are hypothesis-generating.